ceph df max available miscalculation

Turns out max available space is calculated according to the fullest osds in the cluster and has nothing to do with total free space in the cluster. From what i've found this kind of fluctiation mainly happens on small clusters.